In the last week and a half, a lot of computing activity took place for the PPC. We were heavily involved in the Computing Operations Workshop (March 12-13, 2026), the Computing and Offline Week (March 16-20, 2026), and, right in the middle, a CERN seminar on ‘Agents for Scientific Computing’ (recording). The CERN seminar was held by our own Jason Mohoney explaining the status and goals of our Archi project. There were at least four significant talks on Archi, delivered to audiences ranging from internal Computing Operations to CMS global, the LHC experiments, and the entire CERN community.
One might say Archi went viral after we finally released the first production version for the CMS Computing Operations on March 4, 2026.
Another big question was CMS’s plan for the workflow management system. For more than a year, there have been various discussions about replacing WMAgent because it is complex and very difficult to maintain. The Computing and Offline Week, which featured several talks, evaluations, and a long discussion session, unfortunately did not lead to many conclusions. Perhaps the two main technical points are:
- CMS likes its glideInWMS and the integration with HTcondor, and
- the ATLAS workflow management system, Panda, does not really allow a clean integration with glideInWMS and is therefore excluded from further consideration.
One interesting topic, which unfortunately did not receive much follow-up, was the potential impact of modern AI software development tools on this discussion. It seems increasingly clear that tools such as Claude Code, Cursor, and Codex can significantly accelerate development. Dima presented an interesting case study called Condora WMS, in which he built a prototype of a lightweight WMS in just one month, grounding its key ideas in HTCondor and DAGMan. It will be interesting to see how this area evolves and what impact these new AI tools will have not only on software development itself, but also on its long-term maintenance, which represents a large share of the cost of running complex software systems. So, if you ask me: Computing is at a turning point.
